Re: [Evolution] bad separation

[NotZed <notzed helixcode com>]

Ok, this XML discussion has to be one of the weirdest ones i've heard.

First, lets kill a few misassumptions:

Parsing the mbox is very fast.  It is definetly faster, and much more
memory efficient to parse a message at a time, even for the whole mbox
than it is to read in a similar xml file.

We already keep a summary, so we only have to parse the message of interest
anyway - we also store the structure, so at some point we will only
parse the message part of interest too.

Performing a linear search on the headers - which are stored efficiently
in memory - is practical speed wise on modern hardware.  using an
xml tree would hit this because you would have to search for the right
header node first, and decode the entities etc, involving mallocs
and all sorts of penalties.

Keeping the data consistent, and performing some updates are much much
easier with a flat file.

MIME already encapsulates all the structure required, and is easier
to parse and decode than xml is.

XML is a real pain to decode.  Yes, even using a library.  Its slow
and messy, and just plain painful.

Save the bandwidth, xml is absolutely not an appropriate format for storing
this data, it would be a waste even storing the summary info in xml
format.